Co-creation students and AI, Cognitive Trajectory and Epistemic Governance of AI in Higher Education

Eidofüge

Noël Junior Yando Fotso (EPHEC), FOBE Aymeric (EPHEC), NDOMBE Astrale (EPHEC), CLERENS Théo (EPHEC)

Training / education / pedagogy

Open a blank chat in a session. Type a question. Read the answer. Adjust the prompt. Get a better answer. Submit the work. This is how students in higher education interact with generative AI today. Not because their institution deployed it, not because a teacher recommended it, but because it is nearly free, instant, and easy to use. Adoption was not institutional. It was individual, driven by a simple motive: making difficult things easier, and keeping pace with peers doing the same. The tools are provided by private companies (OpenAI, Google, Anthropic, Microsoft) with no educational mandate, no obligation to disclose what their products do to the cognitive development of those who use them in educational settings, and no accountability for the consequences of tools whose adoption is no longer up for debate. What this daily practice produces is a new kind of relationship between a student and their work. The student and the AI co-create. The product exists. But the question no current educational system can answer is this: in that co-creation, who actually thought? What did the student learn? Did they learn anything at all? And is that what we want them to learn? We examine this question through four concrete situations, each naming a distinct dimension of the same problem. Noel is a Health Technologies student at EPHEC. She uses AI to generate proofs of concept, verify clinical data, and speed up literature reviews. Her outputs are solid. Whether she is building the ability to reason on her own about these problems, she does not really know. She has no tool to find out. Neither does her institution. Astrale is an a student in International Business at EPHEC specializing in EU public affairs and lobbying, uses artificial intelligence for an academic assignment on a European directive. Within seconds, the chatbot provides a clear, structured analysis: key actors, power dynamics, and levers of influence. But behind this apparent precision lies a more fragile reality: generic examples, simplified political dynamics, and institutional mechanisms that are sometimes approximate or missing. Yet the discourse remains fluent, convincing, and perceived as neutral. Astrale learns, retains, and integrates without questioning it. Aymeric is a student at EPHEC with the opportunity to do an ERASMUS exchange in China. Under pressure and relatively isolated, he uses AI to manage the stress of his daily coursework and the language barriers involved in academic writing in English. The AI offers constant availability that his supervisory relationship does not. His co-creation is partly emotional. It depends on a tool he did not choose and cannot evaluate. Théo is an international business student at EPHEC who sat an accounting exam where students were allowed to use AI as a support tool. Some of my classmates used it to answer certain questions, but the responses they received sometimes seemed incorrect. Despite their doubts, they trusted the AI more than their own judgment and chose to follow its answers. As a result, several of them made mistakes and ended up failing the exam. This shows that while AI can be helpful, it should always be used with critical thinking. These four situations are not separate problems. They are four faces of the same structural failure: co-creation between students and AI has made the learning process invisible, for students, teachers, institutions, and regulators, at the precise moment when it is most important to see it.

Current policies measure adoption rates, plagiarism cases, and student satisfaction. None of them address the question that should be at the center of every educational policy decision: is the student becoming more independent in their thinking through AI, or is AI gradually replacing their thinking? No country, no institution, and no regulatory framework currently has a tool to answer this question. The risk it hides is structural and generational: a cohort of graduates who perform well with AI and less well without it. A model trained on 0.2% of African data does not produce biased answers by accident. It structurally reproduces the worldview of its training data. Teaching with this tool without naming it reproduces a form of colonialism that does not announce itself. Not the imposition of a language or an administration, but the silent imposition of a cognitive framework presented as neutral. Any institution that deploys AI without addressing its epistemic context will not be providing access to knowledge. It will be providing access to someone else's knowledge, without a label. Regardless of location, university systems put students under pressure and orient them toward research. To produce the tailored answers of the ideal professor they praise, chatbots use user profiles to reduce friction as much as possible. This is genuinely useful when applied well. We ask the question: how would we have discovered that the earth is round with tools that handle language and concepts better than we do? By thinking about these tools and adapting them to work with us. Most AI-driven discoveries today come from exploring logical paths based on concepts already instilled in the model that had simply not yet been followed. The rest are co-creations by people with deep mastery of their subject, who used AI as a pure tool. If students who do not use AI outperform those who do in AI-permitted exams, that is not an argument for banning it. It is a signal that the assessment itself must be reconsidered. But without a mechanism to capture and act on that signal, institutions cannot learn from their own practice. These four problems point to a single failure of governance: institutions certify competences they can no longer observe, in a context they did not choose, with tools they do not control, and without a way to know what those tools are doing to the people in their care.

We are four students, each brought to this challenge from a different angle, but all living the same contradiction: we use AI daily in our studies without a clear framework for understanding what it does to us, and we are each training in our field for the world we will leave to future generations. That world will be shaped directly by the policies we debate today. This event is an opportunity to contribute, at scale, to something larger than ourselves. We would not have missed it for anything. Our shared goal is to contribute something that is not a solution, but a framework: coherent enough to be useful, humble enough to be honest, and open enough to be completed by the communities it is meant to serve.